Neural-Based Adversarial Encryption of Images in ECB Mode with 16-Bit Blocks?

Total Page:16

File Type:pdf, Size:1020Kb

Neural-Based Adversarial Encryption of Images in ECB Mode with 16-Bit Blocks? Neural-Based Adversarial Encryption of Images in ECB Mode with 16-bit Blocks? Pablo Rivas[0000−0002−8690−0987] and Prabuddha Banerjee Computer Science, Marist College, Poughkeepsie NY 12601, USA fpablo.rivas,[email protected] Abstract. Digital images possess rich features that are highly corre- lated among neighboring pixels and highly redundant bits. Thus, the study of encryption algorithms over images is a subject of study to de- termine encryption quality. In this paper, we study the applicability of neural networks in learning to use secret keys to protect information (e.g. images) from other neural networks. We implement a well-known adver- sarial neural network architecture that is designed to learn to establish its own encryption method to protect itself from an attacker that is also a neural network. Our preliminary results suggest that this type of neu- ral architecture is capable of securing communications in the midst of a neural-based attacker using only 16-bit blocks. Keywords: adversarial neural networks · encryption · symmetric key encryption. 1 Introduction Recently, there has been interest in encryption and decryption of communica- tions using Adversarial Neural Networks [1]. This is due to Adversarial Neural Networks which can be trained to protect the communication using Artificial In- telligence. As neural networks can be used for advanced functional specifications which can be categorized into complex task and those are beyond simple func- tional specifications. These objectives include, for example, generating realistic images [5] and solving multiagent problems [3, 9]. Advancing these lines of work, we showed that neural networks can learn to protect their communications in order to satisfy a policy specified in terms of an adversary. Cryptography concerns broadly with algorithms and protocols that will give certainty that the secrecy and integrity of our information is maintained. Cryp- tography is the science of protection our data and communication. A secured mechanism is where it achieves its goal against all attackers. So these form Turing machines or a cryptographic mechanism are formed as pro- grams. Hence attackers may be described in those terms as well as the complex- ity of time and how frequently are they successful in decoding messages. For instance, an encryption algorithm is said to be secure if no attacker can extract ? Supported by the VPAA's Office at Marist College. 2 P. Rivas and P. Banerjee information about plaintexts from ciphertexts. Modern cryptography provides rigorous versions of such definitions [4]. For a given problem Neural networks have the ability to selectively explore a solution space. This feature finds a nat- ural niche of application in the field of cryptanalysis. Neural networks provides new methods of attacking an encryption algorithm based on the fact that neural networks can reproduce any function; this is con- sidered as a very powerful computational tool since any cryptographic algorithm must have an inverse function. In the design of the neural network we study in this paper, adversaries play a very important role. They were originally proposed using what is known as 'adversarial examples' that were found using a neural architecture named a gen- erative adversarial network (GAN) [5]. In this context, the adversaries are neu- ral networks that sample data from a distribution whose parameters are learned from the data. Furthermore, as per the Cryptography definition, practical ap- proaches to training Generative adversarial networks do not consider all possible adversaries, rather adversaries relatively small in size which are improved upon by training. We built our work based upon these ideas. Historically, neural networks have received criticism in the area of cryptog- raphy since earlier models were not able to model an XOR operation, which is commonly used in cryptography. However, nowadays neural networks can easily model XOR operations and be taught to protect the secrecy of data from other neural networks; such models can discover and learn different forms of encryp- tion and decryption, without specifically creating an existing algorithms or even knowing the type of data they will encrypt. In this paper we will discuss the implementation of such architecture in the context of image encryption and will comment on the results obtained. This paper is organized as follows: Section 2 discusses the background pertaining to cryptography and adversarial learning along with the methodology used. Section 3 discusses our experiments and the results obtained. Finally, Section 4 draws conclusions of our paper. 2 Background and Methodology 2.1 Ciphertext-only attacks In the basic architecture proposed by [1], there are 3 neural networks of which the first one is Alice who is part of the cryptographic learning model, so is tasked with encrypting binary strings (e.g. of an image) provided a key. Bob, another friendly neural network, is tasked with learning to decrypt Alice's message. Both Alice and Bob are performing symmetric-key encryption, so they have a common key for encrypting their communication; this key needs to be provided by a human prior to the learning prior to the learning process. Lastly Eve's neural network will be eavesdropping and try to attack or hack their communication by observing the ciphertext. This is known as a ciphertext-only attack [8]. While this is a neural network-based architecture, the overall encryption scheme is well known in textbooks as the one shown in Fig. 1. Neural-Based Adversarial Encryption of Images 3 Fig. 1: A traditional encryption scheme with a twist: Alice, Bob, and Eve are neural networks. 2.2 Block cipher mode: ECB All block ciphers require specific chunks of data (usually binary) of specific sizes that will be used as input or output. Traditionally, these blocks of data are encrypted and decrypted; however, what happens to those blocks internally on the cipher is unique to each cipher. As long as we have information that can be broken down into binary strings of information (e.g. images) we can use any block cipher. In this paper we are concerned with encrypting images. We used the Python programming language in which we take the image and convert into strings of byte arrays and then the byte arrays are flattened and converted into binary strings of the size of the desired block. In this preliminary research we limited our study to 16-bits of information per block. The choice of this block size is due to the large training time for larger blocks. There are several ways or modes in which block ciphers operate. Each mode provides additional security. The only mode that does not provide additional security is known as Electronic Code-Book (ECB) mode. It is important to remark that in order to study the quality of a cipher we must study it by itself with no additional help from other advanced (and much better) modes, so ECB is used here for functionality and to allows us an objective measure of quality of the cipher. The encryption task in ECB mode here is given by: Ci = Ek (Pi) ; i = 1; 2; 3;::: (1) where the function Ek(·) is the encryption algorithm operating with secret key k, Pi is the i-th block of plain text, and Ci is the corresponding i-th block of cipher text. During encryption the key, which is of 16 bits in size, is represented as an array of the following form: 4 P. Rivas and P. Banerjee Fig. 2: ECB mode of operation on an image based on key k. The encrypted image, ciphertext, can be saved, displayed, transmitted, or further analyzed as needed. key = [0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1] Decryption works in similar way but in reverse w.r.t. encryption, i.e., first the cyphertext blocks are passed through the decryption nerual network (Bob) and the original plaintext blocks are recovered. The decryption is here governed by: Pi = Dk (Ci) ; i = 1; 2; 3;::: (2) where Dk(·) is the decryption algorithm operating with key k, and Pi and Ci are as in Eq. (1). The reconstruction process of an image in Python follows a similar process as prior to encryption, i.e. in two stages where in the recovered plaintext is first converted into hexadecimal numbers, then converted to array of bytes, and then these arrays of bytes are thereafter reshaped forming an image. This process is exemplified in Fig. 2. 2.3 Neural architecture Fig. 3 depicts the neural architecture implemented in Python with Keras on Ten- sorFlow; the original architecture was presented in [1], and is based on Generative Adversarial Networks [5]. The neural architecture has two identical networks, Al- ice and Bob; however, Bob is a mirror (reverse) of Alice. The Alice network has a combination of one-dimensional convolutional layers and dense layers; there are a total of four convolutional layers, the first three have hyperbolic tangent activations (tanh), and the last convolutional layer has a sigmoid activation as the output. At the input layer, Alice has a dense layer with a tanh activation. Bob is the exact mirror of Alice, i.e. in opposite order. The input to Alice has to be two vectors: a binary string denoting the plain- text P 2 B16 and a binary string denoting the key k 2 B16. In this implementa- tion we used 16-bits for both. The output of Alice is a vector of codes known as the ciphertext C 2 R16; Neural-Based Adversarial Encryption of Images 5 Fig. 3: Neural architecture of the cipher. During training, Alice and Bob are trained using randomly generated keys and plaintext while Eve attempts to recover the plaintext from the ciphertext and no key. During training, we randomly generated pairs of (P; k) in mini-batches of size 256, which was determined experimentally to be a nice trade-off between training speed and performance.
Recommended publications
  • Exploring Artificial Neural Networks in Cryptography – a Deep Insight
    ISSN 2347 - 3983 Manikandan. N et al., International Journal of EmerginVolumeg Trends 8. No. in 7, Engineering July 2020 Research, 8(7), July 2020, 3811-3823 International Journal of Emerging Trends in Engineering Research Available Online at http://www.warse.org/IJETER/static/pdf/file/ijeter146872020.pdf https://doi.org/10.30534/ijeter/2020/146872020 Exploring Artificial Neural Networks in Cryptography – A Deep Insight Manikandan.N1, Abirami.K2, Muralidharan.D3, Muthaiah.R4 1Research Assistant, School of Computing, SASTRA Deemed to be University, India, [email protected] 2PG Scholar, School of Computing, SASTRA Deemed to be University, India, [email protected] 3Assistant Professor, School of Computing, SASTRA Deemed to be University, India, [email protected] 4Professor, School of Computing, SASTRA Deemed to be University, India, [email protected] securing data/message communication of any cryptosystem. ABSTRACT Cryptanalysis is the art of breaking the security of that cryptosystem [1]. Cryptanalysis might sound controversial, In today’s digital world, cryptography became indispensable but it has its importance, since security property of any in almost all trending technologies. For ensuring a completely cryptosystems is only empirical and hard to measure. Strength secure system with limited computational complexity and efficiency of cryptosystems are directly proportional to involved in the generation of randomness, secret key and the difficulties it involves in breaking the algorithm or other aspects, researchers employed various techniques to the retrieving the secret key [2]. Researches over recent years cryptosystems. Despite the considerable number of strategies, have proved that many cryptographic algorithms are still applications of Artificial neural networks (ANN) for vulnerable to cryptanalytic attacks.
    [Show full text]
  • Performance Comparison Between Deep Learning-Based and Conventional Cryptographic Distinguishers
    Performance comparison between deep learning-based and conventional cryptographic distinguishers Emanuele Bellini1 and Matteo Rossi2 1 Technology Innovation Institute, UAE [email protected] 2 Politecnico di Torino, Italy [email protected] Abstract. While many similarities between Machine Learning and crypt- analysis tasks exists, so far no major result in cryptanalysis has been reached with the aid of Machine Learning techniques. One exception is the recent work of Gohr, presented at Crypto 2019, where for the first time, conventional cryptanalysis was combined with the use of neural networks to build a more efficient distinguisher and, consequently, a key recovery attack on Speck32/64. On the same line, in this work we propose two Deep Learning (DL) based distinguishers against the Tiny Encryption Algorithm (TEA) and its evolution RAIDEN. Both ciphers have twice block and key size compared to Speck32/64. We show how these two distinguishers outperform a conventional statistical distinguisher, with no prior information on the cipher, and a differential distinguisher based on the differential trails presented by Biryukov and Velichkov at FSE 2014. We also present some variations of the DL-based distinguishers, discuss some of their extra features, and propose some directions for future research. Keywords: distinguisher · neural networks · Tiny Encryption Algorithm · differential trails · cryptanalysis 1 Introduction During an invited talk at Asiacrypt 1991 [41], Rivest stated that, since the blossoming of computer science and due to the similarity of their notions and concerns, machine learning (ML) and cryptanalysis can be viewed as sister fields. However, in spite of this similarity, it is hard to find in the literature of the most successful cryptanalytical results the use of ML techniques.
    [Show full text]
  • An Efficient Security System Using Neural Networks
    International Journal of Engineering Science Invention (IJESI) ISSN (Online): 2319 – 6734, ISSN (Print): 2319 – 6726 www.ijesi.org ||Volume 7 Issue 1|| January 2018 || PP. 15-21 An Efficient Security System Using Neural Networks 1*Amit Kore, 2Preeyonuj Boruah, 3Kartik Damania, 4Vaidehi Deshmukh, 5Sneha Jadhav Information Technology, AISSMS Institute of Information Tecchnology, India Corresponding Author: Amit Kore Abstract: Cryptography is a process of protecting information and data from unauthorized access. The goal of any cryptographic system is the access of information among the intended user without any leakage of information to others who may have unauthorized access to it. As technology advances the methods of traditional cryptography becomes more and more complex. There is a high rate of exhaustion of resources while maintaining this type of traditional security systems. Neural Networks provide an alternate way for crafting a security system whose complexity is far less than that of the current cryptographic systems and has been proven to be productive. Both the communicating neural networks receive an identical input vector, generate an output bit and are based on the output bit. The two networks and their weight vectors exhibit novel phenomenon, where the networks synchronize to a state with identical time-dependent weights. Our approach is based on the application of natural noise sources that can include atmospheric noise generating data that we can use to teach our system to approximate the input noise with the aim of generating an output non-linear function. This approach provides the potential for generating an unlimited number of unique Pseudo Random Number Generator (PRNG) that can be used on a one to one basis.
    [Show full text]
  • Neural Synchronization and Cryptography
    Neural Synchronization and Cryptography Π Π Π Dissertation zur Erlangung des naturwissenschaftlichen Doktorgrades der Bayerischen Julius-Maximilians-Universit¨at Wurzburg¨ vorgelegt von Andreas Ruttor arXiv:0711.2411v1 [cond-mat.dis-nn] 15 Nov 2007 aus Wurzburg¨ Wurzburg¨ 2006 Eingereicht am 24.11.2006 bei der Fakult¨at fur¨ Physik und Astronomie 1. Gutachter: Prof. Dr. W. Kinzel 2. Gutachter: Prof. Dr. F. Assaad der Dissertation. 1. Prufer:¨ Prof. Dr. W. Kinzel 2. Prufer:¨ Prof. Dr. F. Assaad 3. Prufer:¨ Prof. Dr. P. Jakob im Promotionskolloquium Tag des Promotionskolloquiums: 18.05.2007 Doktorurkunde ausgeh¨andigt am: 20.07.2007 Abstract Neural networks can synchronize by learning from each other. For that pur- pose they receive common inputs and exchange their outputs. Adjusting discrete weights according to a suitable learning rule then leads to full synchronization in a finite number of steps. It is also possible to train additional neural networks by using the inputs and outputs generated during this process as examples. Several algorithms for both tasks are presented and analyzed. In the case of Tree Parity Machines the dynamics of both processes is driven by attractive and repulsive stochastic forces. Thus it can be described well by models based on random walks, which represent either the weights themselves or order parameters of their distribution. However, synchronization is much faster than learning. This effect is caused by different frequencies of attractive and repulsive steps, as only neural networks interacting with each other are able to skip unsuitable inputs. Scaling laws for the number of steps needed for full syn- chronization and successful learning are derived using analytical models.
    [Show full text]
  • Applications of Search Techniques to Cryptanalysis and the Construction of Cipher Components. James David Mclaughlin Submitted F
    Applications of search techniques to cryptanalysis and the construction of cipher components. James David McLaughlin Submitted for the degree of Doctor of Philosophy (PhD) University of York Department of Computer Science September 2012 2 Abstract In this dissertation, we investigate the ways in which search techniques, and in particular metaheuristic search techniques, can be used in cryptology. We address the design of simple cryptographic components (Boolean functions), before moving on to more complex entities (S-boxes). The emphasis then shifts from the construction of cryptographic arte- facts to the related area of cryptanalysis, in which we first derive non-linear approximations to S-boxes more powerful than the existing linear approximations, and then exploit these in cryptanalytic attacks against the ciphers DES and Serpent. Contents 1 Introduction. 11 1.1 The Structure of this Thesis . 12 2 A brief history of cryptography and cryptanalysis. 14 3 Literature review 20 3.1 Information on various types of block cipher, and a brief description of the Data Encryption Standard. 20 3.1.1 Feistel ciphers . 21 3.1.2 Other types of block cipher . 23 3.1.3 Confusion and diffusion . 24 3.2 Linear cryptanalysis. 26 3.2.1 The attack. 27 3.3 Differential cryptanalysis. 35 3.3.1 The attack. 39 3.3.2 Variants of the differential cryptanalytic attack . 44 3.4 Stream ciphers based on linear feedback shift registers . 48 3.5 A brief introduction to metaheuristics . 52 3.5.1 Hill-climbing . 55 3.5.2 Simulated annealing . 57 3.5.3 Memetic algorithms . 58 3.5.4 Ant algorithms .
    [Show full text]
  • A Review of Machine Learning and Cryptography Applications
    2020 International Conference on Computational Science and Computational Intelligence (CSCI) A Review of Machine Learning and Cryptography Applications Korn Sooksatra∗ and Pablo Rivas†, Senior, IEEE School of Engineering and Computer Science Department of Computer Science, Baylor University ∗Email: Korn [email protected], †Email: Pablo [email protected] Abstract—Adversarially robust neural cryptography deals with its own encryption mechanism. This research captured the the training of a neural-based model using an adversary to attention of many researchers in both areas, cryptography, leverage the learning process in favor of reliability and trustwor- and machine learning [10]–[13]. However, while the authors thiness. The adversary can be a neural network or a strategy guided by a neural network. These mechanisms are proving of [9] gathered much attention, they are other lesser-known successful in finding secure means of data protection. Similarly, successful models that need to be put in the proper context to machine learning benefits significantly from the cryptography display the problems that have been solved and the challenges area by protecting models from being accessible to malicious that still exist. This paper aims to provide context on recent users. This paper is a literature review on the symbiotic rela- research in the intersection between cryptography and machine tionship between machine learning and cryptography. We explain cryptographic algorithms that have been successfully applied in learning. machine learning problems and, also, deep learning algorithms The paper is organized as follows: Section II describes that have been used in cryptography. We pay special attention preliminaries on both deep learning and adversarial neural to the exciting and relatively new area of adversarial robustness.
    [Show full text]
  • Neural Synchronization and Cryptography
    Neural Synchronization and Cryptography Dissertation zur Erlangung des naturwissenschaftlichen Doktorgrades der Bayerischen Julius-Maximilians-Universit¨at Wurzburg¨ vorgelegt von Andreas Ruttor aus Wurzburg¨ Wurzbur¨ g 2006 Eingereicht am 24.11.2006 bei der Fakult¨at fur¨ Physik und Astronomie 1. Gutachter: Prof. Dr. W. Kinzel 2. Gutachter: Prof. Dr. F. Assaad der Dissertation. 1. Prufer:¨ Prof. Dr. W. Kinzel 2. Prufer:¨ Prof. Dr. F. Assaad 3. Prufer:¨ Prof. Dr. P. Jakob im Promotionskolloquium Tag des Promotionskolloquiums: 18.05.2007 Doktorurkunde ausgeh¨andigt am: Abstract Neural networks can synchronize by learning from each other. For that pur- pose they receive common inputs and exchange their outputs. Adjusting discrete weights according to a suitable learning rule then leads to full synchronization in a finite number of steps. It is also possible to train additional neural networks by using the inputs and outputs generated during this process as examples. Several algorithms for both tasks are presented and analyzed. In the case of Tree Parity Machines the dynamics of both processes is driven by attractive and repulsive stochastic forces. Thus it can be described well by models based on random walks, which represent either the weights themselves or order parameters of their distribution. However, synchronization is much faster than learning. This effect is caused by different frequencies of attractive and repulsive steps, as only neural networks interacting with each other are able to skip unsuitable inputs. Scaling laws for the number of steps needed for full syn- chronization and successful learning are derived using analytical models. They indicate that the difference between both processes can be controlled by changing the synaptic depth.
    [Show full text]
  • A Symmetric Key Encryption for Data Integrity Varification Using Artificial Neural Network
    INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 8, ISSUE 10, OCTOBER 2019 ISSN 2277-8616 A Symmetric Key Encryption For Data Integrity Varification Using Artificial Neural Network Monika P, Anita Madona Abstract: An automatic programmed encryption framework dependent on neural networks beginning from the vital symmetric key model, the neural networks to make the conversation of the legal party more effective given less training steps, and to select the suitable hyper-parameters to enhance the statistical randomness to withstand distinguishing attack. In the current neural system, information trustworthiness isn't ensured, and Unique and arbitrary encryption keys are additionally fundamental for data honesty: an open key can't be acquired from an random key generator since it enables mediators to promptly assault the system and access delicate information. In this manner, so as to verify both data and keys, a plan ought to be utilized to deliver particular, mixed and arbitrary middle of the road encryption and decryption keys. Using the Automatic Artificial Neural Network (ANN) to introduce ideas such as authenticated encryption. ANN are the standards of finding the choice consequently by computing the proper parameters (loads) to cause the similarity of the framework and this to can be imperative to have the keys that used in stream figure cryptography to make the general framework goes to high security. The feature learned in Neural Network continues uncertain and we eliminate the chance of being a one-time pad encryption system through the test. With the exception of a few models wherever the assailants are too powerful, most models will be taught to stabilize at the training point.
    [Show full text]
  • An Analysis of Email Encryption Using Neural Cryptography
    Journal of Multidisciplinary Engineering Science and Technology (JMEST) ISSN: 3159-0040 Vol. 2 Issue 1, January - 2015 An Analysis of Email Encryption using Neural Cryptography Sudip Kumar Sahana Prabhat Kumar Mahanti Department of Computer Science & Engineering Department of Computer Science & Applied Stat. Birla Institute of Technology, Mesra University of New Brunswick Ranchi , India Canada [email protected] [email protected] Abstract— Encrypted Email using Neural cryptography, Neural Key Exchange is used to Cryptography is a way of sending and receiving synchronize the two neural networks by encrypted email through public channel. Neural synchronization and Mutual learning is used in the key exchange, which is based on the neural key exchange protocol. The neural key can be synchronization of two tree parity machines, can generated once the network is synchronized. be a secure replacement for this method. This II. METHODOLOGY paper present an implementation of neural cryptography for encrypted E-mail where Tree To increase the confidentiality [7], Encrypted Emails Parity Machines are initialized with random inputs using Neural Cryptography can be used to send and vectors and the generated outputs are used to receive encrypted emails over unsecured channels. train the neural network. Symmetric key The generation of neural key is done through the cryptography is used in encryption and weights assigned to the Tree Parity Machine decryption of messages. [10][13][14][15] either by a server or one of the systems. These inputs are used to calculate the value Keywords— Cryptography; Tree Parity of the hidden neuron and then the output is used to Machine; Neural Key; Encryption; Decryption; predict the output of Tree Parity Machine, as shown in Neurons.
    [Show full text]
  • An Approach to Cryptography Based on Continuous-Variable Quantum
    www.nature.com/scientificreports OPEN An Approach to Cryptography Based on Continuous-Variable Quantum Neural Network Jinjing Shi1,4*, Shuhui Chen1,4, Yuhu Lu1, Yanyan Feng1*, Ronghua Shi1, Yuguang Yang2 & Jian Li3 An efcient cryptography scheme is proposed based on continuous-variable quantum neural network (CV-QNN), in which a specifed CV-QNN model is introduced for designing the quantum cryptography algorithm. It indicates an approach to design a quantum neural cryptosystem which contains the processes of key generation, encryption and decryption. Security analysis demonstrates that our scheme is security. Several simulation experiments are performed on the Strawberry Fields platform for processing the classical data “Quantum Cryptography” with CV-QNN to describe the feasibility of our method. Three sets of representative experiments are presented and the second experimental results confrm that our scheme can correctly and efectively encrypt and decrypt data with the optimal learning rate 8e − 2 regardless of classical or quantum data, and better performance can be achieved with the method of learning rate adaption (where increase factor R1 = 2, decrease factor R2 = 0.8). Indeed, the scheme with learning rate adaption can shorten the encryption and decryption time according to the simulation results presented in Figure 12. It can be considered as a valid quantum cryptography scheme and has a potential application on quantum devices. Cryptography is one of the most crucial aspects for cybersecurity and it is becoming increasingly indispensable in information age. In classical cryptosystems, cryptography algorithms are mostly based on classical hard-to-solve problems in number theory. However, the development of quantum computer and quantum algorithms1,2, such as Shor’s algorithm3, poses an essential threat on the security of cryptosystems based on number theory difcul- ties (like RSA cryptosystem).
    [Show full text]
  • Spiking Neurons with Asnn Based-Methods for The
    International journal of computer science & information Technology (IJCSIT) Vol.2, No.4, August 2010 SPIKING NEURONS WITH ASNN BASED -METHODS FOR THE NEURAL BLOCK CIPHER Saleh Ali K. Al-Omari 1, Putra Sumari 2 1,2 School of Computer Sciences,University Sains Malaysia, 11800 Penang, Malaysia [email protected] , [email protected] ABSTRACT Problem statement: This paper examines Artificial Spiking Neural Network (ASNN) which inter-connects group of artificial neurons that uses a mathematical model with the aid of block cipher. The aim of undertaken this research is to come up with a block cipher where by the keys are randomly generated by ASNN which can then have any variable block length. This will show the private key is kept and do not have to be exchange to the other side of the communication channel so it present a more secure procedure of key scheduling. The process enables for a faster change in encryption keys and a network level encryption to be implemented at a high speed without the headache of factorization. Approach: The block cipher is converted in public cryptosystem and had a low level of vulnerability to attack from brute, and moreover can able to defend against linear attacks since the Artificial Neural Networks (ANN) architecture convey non-linearity to the encryption/decryption procedures. Result: In this paper is present to use the Spiking Neural Networks (SNNs) with spiking neurons as its basic unit. The timing for the SNNs is considered and the output is encoded in 1’s and 0’s depending on the occurrence or not occurrence of spikes as well as the spiking neural networks use a sign function as activation function, and present the weights and the filter coefficients to be adjust, having more degrees of freedom than the classical neural networks.
    [Show full text]
  • A Neural Network Implementation of Chaotic Sequences for Data Encryption
    IJournals: International Journal of Software & Hardware Research in Engineering ISSN-2347-4890 Volume 4 Issue 5 May, 2016 A Neural Network Implementation of Chaotic Sequences for Data Encryption Harshal Shriram Patil Ms.Amita Shrivastava M.Tech Scholar Assistant Professor Astral Institute of Technology and Research, Astral Institute of Technology and Research, Indore, M.P., India Indore, M.P., India [email protected] [email protected] ABSTRACT the need for extremely secure as well as efficient Cryptography has remained as one of the most encryption standards has raised manifold. With sought after fields of research in since the past digital technology becoming accessible to various century. With the advent of digital technology attackers too, the need for secure algorithms for resulting in extremely high data transfer rates, data encryption becomes even more important. newer encryption algorithms which would ensure high level of data security yet have low space and time complexities are being investigated. In this paper we propose a chaotic neural network based encryption mechanism for data encryption. The heart of the proposed work is the existence of chaos in the neural network which makes it practically infeasible to any attacker to predict the output for changing inputs.[1],[2] We illustrate the design of the chaotic network and subsequently the encryption and decryption processes. Finally we examine the performance of the encryption algorithm with standard data encryption algorithms such as DES, AES, Blowfish, Elliptic curve cryptography to prove the efficacy of the proposed algorithm.[1],[2] The parameters chosen for the analysis are throughput, avalanche effect and bit Fig1.
    [Show full text]